杉木全双列杂交试验数据分析方法比较  

Comparison on data analysis methods of complete diallel cross experiment for Chinese fir

作　　者：齐明[1,2] 何贵平 程亚平[3] 肖纪军 王海蓉 邱勇斌 QI Ming;HE Guiping;CHENG Yaping;XIAO Jijun;WANG Hairong;QIU Yongbin(Research Institute of Subtropical Forestry of CAF,Fuyang,Zhejiang 311400,China;Zhejiang Provincial Key Laboratory of Tree Breeding,Fuyang,Zhejiang 311400,China;Qingyuan Forest farm of Qingyuan County,Qingyuan,Zhejiang 323805,China;Foresty Academy of Longquan City,Longquan,Zhejiang 323700,China;Forestry Technology Extension Station of Suichang County,Suichang,Zhejiang 323300,China;Forestry Farm of Kaihua County,Kaihua,Zhejiang 324300,China)

机构地区：[1]中国林业科学研究院亚热带林业研究所,浙江富阳311400 [2]浙江省林木育种技术研究重点实验室,浙江富阳311400 [3]浙江省庆元县庆元林场,浙江庆元323805 [4]浙江省龙泉市林科院,浙江龙泉323700 [5]浙江省遂昌县林业技术推广总站,浙江遂昌323300 [6]浙江省开化县林场,浙江开化324300

出　　处：《福建农林大学学报（自然科学版）》2019年第5期579-584,共6页Journal of Fujian Agriculture and Forestry University:Natural Science Edition

基　　金：浙江省“十三五”林木新品种育种专项资助项目(2016C02056-5);中国林科院中央级公益性科研院所基本科研业务费专项资金资助项目(CAFYBB2017 ZA001-1-2)

摘　　要：为了从杉木全双列交配试验非平衡资料分析中获得一个合适可靠的方差分量,采用不同方法来估计参试因子的方差分量.研究材料是一套杉木6×6全双列(GriffingⅢ)5个不同性状的非平衡资料,分别建立随机模型和混合模型,采用MINQUE(1)法估计方差分量;对于平衡数据采用Yates方法来分析参试因子的方差分量,以转化分析法的结果为参照对象来比较各方法分析结果的一致性和差异性,并权衡各种模型和各种方法分析结果获得的信息量和合理性.比较研究发现:(1)对于GriffingⅢ不平衡试验数据,构建随机模型,采用MINQUE(1)法估计参试因子的方差分量,该法获得的信息多,但分析结果与转化分析法和Yates法的结果有差异,且有负的方差分量;(2)平衡数据Yates法的分析结果与不平衡的试验数据转化分析法的分析结果相似,且具有最小方差和无偏性,还可获得母体细胞质效应的信息.在进行GriffingⅢ不平衡试验数据分析时,建议先采用随机模型的MINQUE(1)来估计方差分量,再采用平衡数据的Yates法分析,两者结果可相互应证.In order to find out a reliable variance component among non-equilibrium data of Chinese fir under full diallel cross trial,random model and mixed model were established based on the data,then variance components of each model were evaluated using MINQUE(1).Meanwhile,the equilibrium data of the diallel cross trial,were retrieved using least square method,with its variance components being evaluated by Yates method.The converted value was used as a benchmark to select a proper model with regards to the quantity and accuracy of information.Results showed that random model resulted in higher quantity of information but also nega-tive variance component and higher disparity from the reference value.Having minimum variance and good unbiasedness,results of Yates method shared a similar trend with the transformation analysis,and also showed a maternal cytoplasmic effect.Therefore,when it comes to the analysis on non-equilibrium data under GriffingⅢ,it is suggested to start with MINQUE(1)of random model on the analysis of variance,and then use Yates method with equilibrium data,so that the results of both approaches can be mutually verified.

关 键 词：随机模型 混合模型 不平衡数据 MINQUE(1) Yates法 

分 类 号：S718.46[农业科学—林学]